pt-kill shouldn't check if STDIN is a tty when --daemonize is given

[William Shallum]

I'm trying to run pt-kill from monit. STDIN is not a tty then. The problem is that pt-kill assumes it is going to read processlist from STDIN and I can find no way of disabling that assumption.

I don't think anyone would actually pipe a processlist to a process that is going to be daemonized so wouldn't it be better to check the --daemonize flag first before checking that STDIN is a tty?

Percona Toolkit version: 1.0.1

Thanks,
William

[James Byers (jbyers)]

I believe this also affects pt-kill when run from cron. I'd be happy to supply a patch, but I'm not sure which behavior is intended. The docs say that '-' specified as a file trigger reading STDIN, however the last usage example shows STDIN detection:

  mysql -e "SHOW PROCESSLIST" | pt-kill --busy-time 60 --print

My personal preference would be for pt-kill to require '-' when used in this way.

[Daniel Nichter (daniel-nichter)]

pt-kill really only reads files on the command line (or piped via STDIN) for testing purposes. Normally, it connects to MySQL directly to get the processlist. How are you running the tool? It should be something like: pt-kill <options> --host <host> --user <user> --password <pass>

[Baron Schwartz (baron-xaprb)]

This tool isn't very user-friendly and it's one that I've got on my list to re-spec for a future release. I've had the same problem as the OP did.

[Daniel Nichter (daniel-nichter)]

This problem came up elsewhere, which lead to a discussion, and short story: the reporter is correct. On my Mac, for example, cron apparently pipes something to STDIN which is *not* a TTY, so this code

   if ( !-t STDIN ) {
      PTDEBUG && _d("STDIN is piped");
      @ARGV = ('-');
   }

makes the tool read from STDIN, but there's nothing there, so it exits immediately even though --daemonize was given. Reading from STDIN was so that we could test matching like "cat proclist | pt-kill ...". This is unreliable though because, for another example, on Ubuntu, cron also pipes something to STDIN but Perl says it *is* a TTY so the tool thinks it's running "normally" and therefore connects to MySQL as expected.

In short, testing STDIN and testing via STDIN are problematic so we're removing this and adding the option --test-matching FILES to make it explicit when we (or the user perhaps) wants to read FILES to test matching.

[Daniel Nichter (daniel-nichter)]

This bug was not released in 2.0.3 as originally planned because the branch wasn't merged. It will be released in 2.0.4.

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PT-437